期刊名称：International Journal of Computing and Business Research
电子版ISSN：2229-6166
出版年度：2013
卷号：4
期号：1
出版社：International Journal of Computing and Business Research
摘要：Data mining (the analysis step of the "Knowledge Discovery in Databases" process, or KDD), is the process that results in the discovery of new patterns in large data sets. Classification is a data mining technique used to map a data item into one of several predefined classes. There are many classification methods to classify instances, but we don’t know which classification method is suitable for our dataset i.e. which classification algorithm will give less error. This article evaluates the performance of different classification techniques and compares them, based on the parameter – “Mean Absolute Error”. Classification methods covered in this work include Bayesian Networks, Neural Networks, Support Vector Machines, and Nearest Neighbor. To render more credibility to the results, the target algorithms have been tested on five datasets taken from UCI Machine Learning Repository. This comparison will show which algorithm is best, in terms of mean absolute error i.e. which will give less error. The performance of classification techniques are evaluated by using open source software named “WEKA” (Waikato Environment for Knowledge Analysis).
关键词：Data mining (the analysis step of the "Knowledge;Discovery in Databases" process; or KDD); is the process;that results in the discovery of new patterns in large data;sets. Classification is a data mining technique used to map a;data item into one of several predefined classes. There are;many classification methods to classify instances; but we;don’t know which classification method is suitable for our;dataset i.e. which classification algorithm will give less error.;This article evaluates the performance of different;classification techniques and compares them; based on the;parameter – “Mean Absolute Error”. Classification methods;covered in this work include Bayesian Networks; Neural;Networks; Support Vector Machines; and Nearest Neighbor.;To render more credibility to the results; the target;algorithms have been tested on five datasets taken from UCI;Machine Learning Repository. This comparison will show;which algorithm is best; in terms of mean absolute error i.e.;which will give less error. The performance of classification;techniques are evaluated by using open source software;named “WEKA” (Waikato Environment for Knowledge;Analysis).
